#include <mpi.h>
#include <stdlib.h>
#include <stdio.h>

int main(int argc, char *argv[])
{
    int rank, nproc;
    int row1,col1;
    int row2,col2;
    FILE *fp1, *fp2, *fp3;
    unsigned int i,j,k,sum;
    int *matrix1, *matrix2, *matrixtemp, *temp1, *temp2, *matrix1r, *matrix2r, *result;
    double StartTime, EndTime;

    MPI_Init(&argc, &argv);
    MPI_Comm_rank(MPI_COMM_WORLD, &rank);
    MPI_Comm_size(MPI_COMM_WORLD, &nproc);

    if(rank==0){
        fp1 = fopen("/Users/win/Desktop/MatrixMultiplication_large/matAlarge.txt", "r");
        fp2 = fopen("/Users/win/Desktop/MatrixMultiplication_large/matBlarge.txt", "r");
        //fp1 = fopen("/Users/win/Desktop/MatrixMul/matrix2.txt", "r");
        //fp2 = fopen("/Users/win/Desktop/MatrixMul/matrix2.txt", "r");
        fp3 = fopen("/Users/win/Desktop/matSol.txt", "w");

        fscanf(fp1, "%d", &row1);
        fscanf(fp1, "%d", &col1);
        fscanf(fp2, "%d", &row2);
        fscanf(fp2, "%d", &col2);

        fprintf(fp3, "%d %d\n", row1, col2);

        MPI_Bcast(&row1, 1, MPI_INT, 0, MPI_COMM_WORLD);
        MPI_Bcast(&col1, 1, MPI_INT, 0, MPI_COMM_WORLD);
        MPI_Bcast(&row2, 1, MPI_INT, 0, MPI_COMM_WORLD);
        MPI_Bcast(&col2, 1, MPI_INT, 0, MPI_COMM_WORLD);

        matrix1 = (int*)malloc(row1 * col1 * sizeof(int));
        matrix2 = (int*)malloc(row2 * col2 * sizeof(int));
        temp1 = (int*)malloc(row1 * col1 * sizeof(int));
        temp2 = (int*)malloc(row2 * col2 * sizeof(int));
        result = (int*)malloc(row1 * col2 * sizeof(int));

        for(i=0; i<row1; i++){
            for(j=0; j<col1; j++){
                fscanf(fp1, "%d", &matrix1[(i*col1)+j]);
            }
        }

        //transpose matrix
        for(i=0; i<row2; i++){
            for(j=0; j<col2; j++){
                fscanf(fp2, "%d", &matrix2[(j*row2)+i]);
            }
        }

        StartTime = MPI_Wtime();

        for(i=1;i<nproc;i++){
            if(i==nproc-1){
                MPI_Send(&matrix1[i*(row1/nproc)*col1], (row1/nproc)*col1 + (row1%(row1/nproc))*col1, MPI_INT, i, 1, MPI_COMM_WORLD);
            } else {
                MPI_Send(&matrix1[i*(row1/nproc)*col1], (row1/nproc)*col1, MPI_INT, i, 1, MPI_COMM_WORLD);
            }
            MPI_Send(&matrix2[0], row2*col2, MPI_INT, i, 2, MPI_COMM_WORLD);
        }

        for(i=0;i<row1/nproc;i++){ // loop row matrix1
            for(j=0;j<col2;j++){ // loop row matrix2
                sum = 0;
                for(k=0;k<row2;k++){ // loop col matrix2 (sum)
                    sum += matrix1[(i*col1)+k] * matrix2[(j*row2)+k];
                }
                result[(i*col2)+j] = sum;

            }
        }

        for(i=1;i<nproc;i++){
            if(i==nproc-1){
                MPI_Recv(&result[i*(row1/nproc)*col2], (row1/nproc)*col2  + (row1%(row1/nproc))*col2, MPI_INT, i, 1, MPI_COMM_WORLD, MPI_STATUS_IGNORE);
            } else {
                MPI_Recv(&result[(i*(row1/nproc))*col2], (row1/nproc)*col2, MPI_INT, i, 1, MPI_COMM_WORLD, MPI_STATUS_IGNORE);
           }
        }

        EndTime = MPI_Wtime();

        for(i=0; i<row1; i++){
            for(j=0; j<col2; j++){
                fprintf(fp3, "%d ", result[(i*col2)+j]);
            }
            fprintf(fp3, "\n");
        }

        fclose(fp1);
        fclose(fp2);
        fclose(fp3);
        printf("Timings : %f Sec\n", EndTime - StartTime);
    } else {

        MPI_Bcast(&row1, 1, MPI_INT, 0, MPI_COMM_WORLD);
        MPI_Bcast(&col1, 1, MPI_INT, 0, MPI_COMM_WORLD);
        MPI_Bcast(&row2, 1, MPI_INT, 0, MPI_COMM_WORLD);
        MPI_Bcast(&col2, 1, MPI_INT, 0, MPI_COMM_WORLD);

        matrix1r = malloc(row1 * col1 * sizeof(int));
        matrix2r = malloc(row2 * col2 * sizeof(int));
        matrixtemp = malloc(row1 * col2 * sizeof(int));
        if(rank==nproc-1){
           MPI_Recv(&matrix1r[0], (row1/nproc)*col1 + (row1%(row1/nproc))*col1, MPI_INT, 0, 1, MPI_COMM_WORLD, MPI_STATUS_IGNORE);
        } else {
        MPI_Recv(&matrix1r[0], (row1/nproc)*col1, MPI_INT, 0, 1, MPI_COMM_WORLD, MPI_STATUS_IGNORE);
        }
        MPI_Recv(&matrix2r[0], row2*col2, MPI_INT, 0, 2, MPI_COMM_WORLD, MPI_STATUS_IGNORE);


        if(rank==nproc-1){
            for(i=0;i<(row1/nproc) + (row1%(row1/nproc));i++){ // loop row matrix1
                for(j=0;j<col2;j++){ // loop row matrix2
                    sum = 0;
                    for(k=0;k<row2;k++){ // loop col matrix2 (sum)
                        sum += matrix1r[(i*col1)+k] * matrix2r[(j*row2)+k];
                    }
                    matrixtemp[(i*col2)+j] = sum;
                }
            }
        } else {
        for(i=0;i<row1/nproc;i++){ // loop row matrix1
            for(j=0;j<col2;j++){ // loop row matrix2
                sum = 0;
                for(k=0;k<row2;k++){ // loop col matrix2 (sum)
                    sum += matrix1r[(i*col1)+k] * matrix2r[(j*row2)+k];
                    }
                    matrixtemp[(i*col2)+j] = sum;

                }
            }
        }

        printf("Process %d done!\n", rank) ;
        if(rank==nproc-1){
           MPI_Send(&matrixtemp[0], (row1/nproc)*col2 + (row1%(row1/nproc))*col2, MPI_INT, 0, 1, MPI_COMM_WORLD);
        } else {
        MPI_Send(&matrixtemp[0], (row1/nproc)*col2, MPI_INT, 0, 1, MPI_COMM_WORLD);
        }
    }

    MPI_Finalize();

    return 0;
}